Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

HGraphScale: Hierarchical Graph Learning for Autoscaling Microservice Applications in Container-based Cloud Computing

Zhengxin Fang, Hui Ma, Gang Chen, Rajkumar Buyya

TL;DR

This work addresses autoscaling of microservice applications in container-based clouds where complex dependencies and deployment schemes challenge existing approaches. It introduces HGraphScale, which combines a cloud-oriented Hierarchical Graph Neural Network (CHGNN) with evolutionary reinforcement learning to learn container embeddings from a three-layer PM–VM–container graph and to produce scaling actions via a two-MLP policy network. Key contributions include (i) a three-layer hierarchical graph representation of the cloud, (ii) CHGNN with bottom-up aggregation that captures global context for container-level decisions, (iii) a scaling policy network and a DRL training regime (ERL) to optimize latency under budget with a capacity-based load balancer, and (iv) extensive experiments showing substantial ART reductions (up to 80.16%) while respecting budgets and providing robust tail latency performance. The results indicate that HGraphScale offers a principled, scalable autoscaling solution for dynamic microservice workloads, with practical impact on QoS and cost efficiency in cloud environments.

Abstract

Microservice architecture has become a dominant paradigm in application development due to its advantages of being lightweight, flexible, and resilient. Deploying microservice applications in the container-based cloud enables fine-grained elastic resource allocation. Autoscaling is an effective approach to dynamically adjust the resource provisioned to containers. However, the intricate microservice dependencies and the deployment scheme of the container-based cloud bring extra challenges of resource scaling. This article proposes a novel autoscaling approach named HGraphScale. In particular, HGraphScale captures microservice dependencies and the deployment scheme by a newly designed hierarchical graph neural network, and makes effective scaling actions for rapidly changing user requests workloads. Extensive experiments based on real-world traces of user requests are conducted to evaluate the effectiveness of HGraphScale. The experiment results show that the HGraphScale outperforms existing state-of-the-art autoscaling approaches by reducing at most 80.16\% of the average response time under a certain VM rental budget of application providers.

HGraphScale: Hierarchical Graph Learning for Autoscaling Microservice Applications in Container-based Cloud Computing

TL;DR

This work addresses autoscaling of microservice applications in container-based clouds where complex dependencies and deployment schemes challenge existing approaches. It introduces HGraphScale, which combines a cloud-oriented Hierarchical Graph Neural Network (CHGNN) with evolutionary reinforcement learning to learn container embeddings from a three-layer PM–VM–container graph and to produce scaling actions via a two-MLP policy network. Key contributions include (i) a three-layer hierarchical graph representation of the cloud, (ii) CHGNN with bottom-up aggregation that captures global context for container-level decisions, (iii) a scaling policy network and a DRL training regime (ERL) to optimize latency under budget with a capacity-based load balancer, and (iv) extensive experiments showing substantial ART reductions (up to 80.16%) while respecting budgets and providing robust tail latency performance. The results indicate that HGraphScale offers a principled, scalable autoscaling solution for dynamic microservice workloads, with practical impact on QoS and cost efficiency in cloud environments.

Abstract

Microservice architecture has become a dominant paradigm in application development due to its advantages of being lightweight, flexible, and resilient. Deploying microservice applications in the container-based cloud enables fine-grained elastic resource allocation. Autoscaling is an effective approach to dynamically adjust the resource provisioned to containers. However, the intricate microservice dependencies and the deployment scheme of the container-based cloud bring extra challenges of resource scaling. This article proposes a novel autoscaling approach named HGraphScale. In particular, HGraphScale captures microservice dependencies and the deployment scheme by a newly designed hierarchical graph neural network, and makes effective scaling actions for rapidly changing user requests workloads. Extensive experiments based on real-world traces of user requests are conducted to evaluate the effectiveness of HGraphScale. The experiment results show that the HGraphScale outperforms existing state-of-the-art autoscaling approaches by reducing at most 80.16\% of the average response time under a certain VM rental budget of application providers.

Paper Structure

This paper contains 32 sections, 16 equations, 13 figures, 6 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Heuristic-based Autoscaling
  4. Reinforcement Learning-based Autoscaling
  5. Graph Neural Network-based Autoscaling
  6. Summary
  7. Problem Description
  8. Proposed Autoscaling Approach
  9. RL Formulation
  10. State
  11. Action
  12. Optimization objective
  13. Hierarchical Graph Representation
  14. Cloud-Oriented Hierarchical Graph Neural Network
  15. Machine embedding Learning
  16. ...and 17 more sections

Figures (13)

  • Figure 1: Microservice application deployed in the container-based cloud with fluctuating user requests.
  • Figure 2: Microservice application deployed in the container-based cloud
  • Figure 3: The overall framework of HGraphScale for the AMC problem.
  • Figure 4: An example of the hierarchical graph representation of the container-cloud.
  • Figure 5: The architecture of the CHGNN.
  • ...and 8 more figures